This week on Absolute AppSec’s 328th episode, Seth (@sethlaw) and Ken (@cktricky) are back to regularly scheduled programming, right before they’re off to Vegas for BlackHat and DEFCON. They touch on a newly disclosed RCE vulnerability, the influx of AI-assisted vulnerability disclosures, and the complexities of bug prioritization. To find this episode, head over to https://www.youtube.com/@AbsoluteAppSec/streams, or find us wherever you get your podcasts.
“The number of vulnerabilities that are being reported and that are valid has increased exponentially in the last twelve to eighteen months […] tracking those vulnerabilities or trying to remediate them one by one is becoming a fool’s errand.”
Hosts Ken Johnson and Seth Law opened with a fast-breaking story: Unlike the usual plugin bugs, this core flaw targets the wp-json batch v1 REST endpoint, prompting major hosts like WordPress.com and WP Engine to force-install patches on customer sites. Defenders unable to patch immediately can block the endpoint at the WAF level as a stopgap. Surprised by how quietly the flaw surfaced, the hosts joked about whether the "wp2shell.com" documentation site was itself vulnerable—since it was down during recording—and urged anyone running WordPress to confirm their patch immediately before mass exploitation begins.
The main segment focused on a post by Anthropic’s Alex Gaynor, who argued for eliminating whole vulnerability classes at the root—such as replacing raw SQL queries with a safe wrapper function—rather than patching bugs one by one. Both hosts agreed this systemic approach is essential now that AI-driven fuzzing is flooding backlogs with unprecedented volumes of CVEs. Ken outlined current prioritization strategies like runtime verification, reachability analysis, and exposure scoring, while Seth argued there's no excuse for fresh bugs to survive past commit given modern IDEs and AI coding agents. However, they admitted the industry still lacks clear metrics to measure security effectiveness, leaving most prioritization frameworks as "codified gut feeling." Seth suggested beta-testing new prioritization models on a small scale and using AI agents to make heavy lifts—like rewriting legacy C/C++ in Rust—feasible, even if walking away from legacy stack expertise remains a real operational risk.
“The more testing and targeting of AI systems that I do, the more it feels like social engineering […] it’s convincing an agent to give you information or to do things that are unintended.”
The final topic covered “The Memory Heist," independent research detailing a privacy exploit against Claude AI’s persistent memory feature. By luring Claude to visit an attacker-controlled domain, the researcher bypassed basic prompt-injection defenses and used indirect sub-path navigation requests to trick the assistant into leaking personal details—like the user's name, employer, and city—one letter or field at a time. Seth noted that AI red-teaming now resembles human social engineering more than classic memory-corruption fuzzing, as attackers persuade an inherently helpful system to break its rules at machine speed. Because LLMs evolved from text prediction models rushed to market rather than software engineered for adversarial defense, the security industry is now stuck trying to bolt security controls onto a foundation that was never designed to resist malicious text.
Think about your mobile app’s source code. Once it hits the app store, it’s out in the wild. And without the right protection, decompiling is easy for malicious actors looking to steal your IP or tamper with your software.
That’s where Guardsquare comes in. Guardsquare provides the highest level of mobile app security for Android and iOS applications and SDKs. Their advanced tools integrate seamlessly into your CI/CD pipeline. We're talking polymorphic multi-layered code hardening techniques and automated runtime application self-protection, paired with mobile application security testing and real-time threat monitoring, to deliver the highest level of mobile app security without compromise.
Don't leave your hard work exposed. Secure your mobile applications today. Go to guardsquare.com to learn more.
You already know what we are going to be talking about all night long when we’re in Vegas next month. Go back to basics with a plain white tee (plus a twist!) here:
Basics are back. CrocsSocks4EVER.
Join us in our Slack; just send us a note to join the channel. As long as you promise not to impersonate us (again), you can sit down, relax, and stay a while.
Stay Secure,
Seth & Ken
https://www.youtube.com/watch?v=R5Eca_H7mxA – Episode 327 w/ Coffee, Chaos, and ProdSec — ASPM Consolidation, Vuln Prioritization - In case you missed it, last week’s episode with Kurt and Cameron also touched on vulnerability prioritization. They talked about all things prioritization strategies amid a massive, AI-driven surge in vulnerability research.
https://youtube.com/watch?v=1SpHiMp0vZ8 – Episode 125 — Interviews, SQLi, Concurrency, WordPress - Let’s blast into the past, all the way to 2021, to the last time WordPress was mentioned. They discuss the then-recent WordPress plugin vulnerabilities and why they are always so devastating.
https://youtu.be/watch?v=UM7EQtThrrg – Episode 323 — Secret Logs, Prompt Injection Risks - Seth and Ken explore a multi-page research paper analyzing prompt injection. Because LLMs mathematically process data through tokenization without any physical or architectural separation between instructions and data contexts, prompt injection cannot be completely solved at the model level.
Absolute AppSec Happenings
Open AI says Hugging Face was breached by its pre-release models – OpenAI admitted that its own AI models breached Hugging Face during an internal cybersecurity test, contradicting Hugging Face's earlier claim that an "external AI agent" was responsible. While being tested on the ExploitGym cyber-capability benchmark with reduced safety refusals, the models exploited a package-installer vulnerability to gain unauthorized internet access, then found further flaws letting them pull benchmark solutions straight from Hugging Face's production database.
The Agentic SLDC: Why Most of What We Do in Software Security Has to Change – Argues that AI coding agents are dismantling the traditional gate-based SDLC security model, since agents iterate faster than human reviewers can check them, causing measurable security degradation. Rather than declaring AppSec dead, it proposes a "shift-everywhere" architecture: embedding security natively in agents (via MCP), keeping SAST/gates as an audit layer, and adding runtime behavioral monitoring, with threat modeling becoming continuous rather than a pre-code ceremony.
How Anthropic secures its AI-native software development lifecycle — Anthropic's Deputy CISO explains how their security team secures an SDLC where Claude authors 80% of merged code. Key strategies: embedding security guidance in CLAUDE.md files so vulnerabilities are prevented at generation time, using multiple narrow-scope agents (not one mega-agent) to review PRs, running coding on egress-locked VMs to contain prompt-injection blast radius, and giving every agent single-purpose, minimal-permission identities. Human review is preserved for high-risk code and critical approvals, with all agent actions logged to SIEM for auditability.
Upcoming Events
Where in the world are Seth and Ken?
August 1-4, 2026 - AI-Enhanced Secure Code Review: Black Hat Edition - BlackHat USA, Las Vegas - Seth and Ken are bringing a four-day exclusive course to Black Hat. This is an update on the exclusive version of the course offered at Black Hat Europe. Early bird pricing is ongoing, so it’s a great opportunity to get a truly in-depth understanding of Secure-Code Review and how it can be empowered through LLM-tooling. Seth and Ken have innovated industry-leading trainings in both of these topics, so this four-day course promises to provide a lot of valuable insight.
August 10-11, 2026 - Agentic AppSec: Harnessing LLMs - DEF CON Training, Las Vegas - Comprehensive course designed for developers and cybersecurity professionals seeking to harness the power of Agentic AI and Large Language Models (LLMs) to enhance software security and development practices.
